An enterprise team wants a managed offering that both provides access to large language models and includes built-in services to adapt models to specific use cases. According to the guidance, which layer of the Generative AI application stack provides this?